Automated Method for Modifying Neural Dynamics

1. A method for improving neural dynamics of an artificial neural network stored in a device, comprising: obtaining prototypical neuron dynamics from a prototypical neuron, the neuron dynamics comprising at least spike timing; and modifying parameters of a piecewise linear neuron model, of the artificial neural network, so that at least a spike timing of the piecewise linear neuron model matches at least a spike timing of the prototypical neuron, the parameters being modified based on an objective function that determines a metric that quantizes a difference between a first function of spike times of the piecewise linear neuron model and a time constant and a second function of spike times of the prototypical neuron and the time constant.

2. The method of claim 1 , in which the objective function minimizes a required number of bits in a representation of the piecewise linear neuron model.

3. The method of claim 1 , in which the objective function reduces model complexity.

4. An artificial neural network, the artificial neural network comprising: a memory unit; and at least one processor coupled to the memory unit, the at least one processor being configured: to obtain prototypical neuron dynamics from a prototypical neuron, the neuron dynamics comprising at least spike timing; and to modify parameters of a piecewise linear neuron model, of the artificial neural network, so that at least a spike timing of the piecewise linear neuron model matches at least a spike timing of the prototypical neuron, the parameters being modified based on an objective function that determines a metric that quantizes a difference between a first function of spike times of the piecewise linear neuron model and a time constant and a second function of spike times of the prototypical neuron and the time constant.

5. The artificial neural network of claim 4 , in which the objective function minimizes a required number of bits in a representation of the piecewise linear neuron model.

6. The artificial neural network of claim 4 , in which the objective function reduces model complexity.

7. An apparatus for improving neural dynamics, comprising: means for obtaining prototypical neuron dynamics from a prototypical neuron, the neuron dynamics comprising at least spike timing; and means for modifying parameters of a piecewise linear neuron model, of the artificial neural network, so that at least a spike timing of the piecewise linear neuron model matches at least a spike timing of the prototypical neuron, the parameters being modified based on an objective function that determines a metric that quantizes a difference between a first function of spike times of the piecewise linear neuron model and a time constant and a second function of spike times of the prototypical neuron and the time constant.

8. A non-transitory computer-readable medium having program code recorded thereon, the program code comprising: program code to obtain prototypical neuron dynamics from a prototypical neuron stored in a memory unit, the neuron dynamics comprising at least spike timing; and program code to modify parameters of a piecewise linear neuron model, of the artificial neural network, so that at least a spike timing of the piecewise linear neuron model matches at least a spike timing of the prototypical neuron, the parameters being modified based on an objective function that determines a metric that quantizes a difference between a first function of spike times of the piecewise linear neuron model and a time constant and a second function of spike times of the prototypical neuron and the time constant.